Welsh Ivan D, Allison Jane R
Centre for Theoretical Chemistry and Physics, Institute of Natural and Mathematical Sciences, Massey Unversity, Private Bag 102904, Auckland, New Zealand.
School of Biological Sciences, University of Auckland, Private Bag 92019, Auckland, New Zealand.
J Cheminform. 2019 Mar 6;11(1):18. doi: 10.1186/s13321-019-0340-0.
Bond orders and formal charges are fundamental chemical descriptors. In cheminformatic applications it is necessary to be able to assign these properties to a given molecular structure automatically, given minimal input information. Here we describe a method for determining the bond order and formal charge assignments from only the atom types and connectivity. Our method utilises a graph theoretical description of electron positions. Each electron position assignment is scored according to lookup tables of atomic and bond dissociation energies derived from quantum chemical calculations. We tested three different optimisation methods-local optimisation, an A* pathfinding method, and an FPT optimisation method utilising tree decompositions-for finding the best electron position assignment, from which the bond orders and formal charges are extracted. We show that our method can assign bond orders and formal charges at a high degree of accuracy across a wide range of molecules from two different databases, and that the FPT algorithm provides the best combination of speed and accuracy.
键级和形式电荷是基本的化学描述符。在化学信息学应用中,鉴于输入信息最少,有必要能够自动将这些属性分配给给定的分子结构。在此,我们描述一种仅根据原子类型和连接性来确定键级和形式电荷分配的方法。我们的方法利用了电子位置的图论描述。根据从量子化学计算得出的原子和键解离能查找表,对每个电子位置分配进行评分。我们测试了三种不同的优化方法——局部优化、A*寻路方法以及利用树分解的固定参数可解(FPT)优化方法——以找到最佳电子位置分配,从中提取键级和形式电荷。我们表明,我们的方法能够在来自两个不同数据库的广泛分子范围内高精度地分配键级和形式电荷,并且FPT算法提供了速度和准确性的最佳组合。